Luma AI, an artificial intelligence company based in Palo Alto, unveiled its new video generation model, Ray2 Flash, last week. This model is a streamlined iteration of the recently launched Ray 2 frontier video generation model. The company claims that Ray2 Flash is faster and more cost-effective than its predecessor while maintaining the ability to produce realistic video content. This multimodal model accepts both text and image inputs and is accessible through the Dream Machine platform, which debuted last year.
Luma AI Introduces Ray2 Flash AI Model
The official Luma AI account made the announcement through a post on X (formerly Twitter). Currently, the Ray2 Flash model is available on the Dream Machine platform, with an application programming interface expected to be launched soon.
According to the announcement, the new video generation model operates at three times the speed and at one-third the cost of the Ray2 frontier model. Additionally, it offers unlimited access, allowing users to create more videos using the same credit allocation. However, Ray2 Flash is exclusively available to paying subscribers of the Dream Machine service.
For free-tier users, a one-time allocation of 400 credits is provided. Following the completion of these credits, users must subscribe to continue using the service. The monthly membership for the platform starts at $6.99 (approximately Rs. 609) and provides 3,200 monthly credits.
The Ray2 Flash model builds upon the capabilities of the Ray2 frontier model, which was launched in January. At that time, Luma AI highlighted the model’s competence in generating realistic videos with natural motion and coherence, alongside a strong adherence to user prompts.
The Dream Machine platform was introduced last year, showcasing both image and video generation capabilities. Users can produce outputs in various styles including cinematic, animation, and realistic formats. However, a trial by Gadgets 360 indicated that while the platform performs well, it struggles with complex prompts or multiple character scenarios, resulting in sub-par prompt adherence.
